Chlorophyll Discrimination in an After Bloom Condition in Inland Waters Using Field, MERIS and Simulated OLCI Data
Streher, Annia Susin; Ferreira, Rafael Damiati; Barbosa, Claudio Clemente
National Institute for Space Research, BRAZIL

Empirical relationships between surface reflectance and optically active substances (OAS) are often used as basis to produce algorithms to estimate the OAS from remote sensing data. The discrimination of aquatic environments with different chlorophyll-a concentrations may be useful for the management of water resources, optimizing eutrophication control in reservoirs. This study aims to evaluate the potential of different spectral data on the chlorophyll-a discrimination during an after bloom condition in a tropical reservoir. Ibitinga reservoir is located in the central region of the state of São Paulo (Brazil), within the Tietê River basin. Ibitinga is a eutrophic reservoir subjected a phytoplankton blooms occurring seasonally in the month of October. Field data collection was carried out on October 23 and 24, 2011. During these two days, nine water stations along the reservoir were sampled. Chlorophyll-a was extracted from samples using 80% ethanol at a temperature of 75°C, left to cool in the dark for 6 to 24h, and then quantified by spectrophotometer analysis. In each sampling station, field spectra were collected using the ASD FieldSpec HandHeld, from 325 nm to 1075 nm, with 1nm of spectral resolution. The OLCI bands were simulated using the average of the field spectra based on the width of the OLCI bands, provided by ESA. The MERIS image was acquired throughout the ESA project ID10990, and refers to March of 2012, which was the nearest date available to the field campaign. The MERIS image was atmospheric corrected with the SMACC algorithm. The relationship between chlorophyll concentration and spectral reflectance was evaluated using the Pearson correlation coefficient, for all available bands. Once correlations were calculated, the two bands with the highest coefficients (band x and band y) were combined using a simple ratio, and the resulting values used to fit a simple linear regression model to estimate mean chlorophyll concentration. The same process was used to find bands for the three band model. For the two band model, the best combination of bands were 677 and 694 nm for the field data, 665 and 705 nm for the MERIS image and 673.5 and 708.75 nm for the OLCI simulated data. The band ratio for the field spectra presented a strong correlation with the chlorophyll, with a R2 = 0.83. As for the MERIS image and the OLCI simulated data the results weren't satisfactory, with R2= 0.25 and R2= 0.13, respectively. The three bands regression model showed better results when compared to the two bands algorithm. The bands used in this model were centered in 620 nm, 665 nm and 900 nm to the MERIS image and in 681.2, 708.75, 767.5 to the OLCI simulated data. The determination coefficient were R2 =0.53 e R2= 0.55, respectively. The difference between the coefficients encountered may be related to a few factors, such as the spectral resolution difference between the data analyzed. The coefficients found for the OLCI simulated data suggests that the inclusion of a third band in the model reduce the interference of the suspended solids in the spectral signal. For the MERIS image, the scale difference between the data acquisition may interfere in the models adjustments, considering that the image pixel integrates a 300 square meters area. For Ibitinga reservoir, a few pixels were enough to cover the open water area. The extraction difficulty of pure water pixels maximize the influence of aquatic and riparian vegetation in the spectral signal, resulting in the selection of the 900nm centered band to the three band model. Another factor to take into account is the atmosphere interference in the orbital data when compared to the in situ, even though the MERIS image was atmospherically corrected. Furthermore, it's necessary to highlight the phytoplankton physiological state during the sampling process. The senescence process of those organisms results in less light absorption by chlorophyll, reducing the differences in the spectral features (depth and width) as the pigment concentrations varies. The present study indicates that the new OLCI sensor represents a potential tool for inland tropical water bodies monitoring, even in after bloom conditions, emphasized by the three band model results.